@cypeg: We have many players here that play only their group of friends (and there is nothing wrong with this). But in each group there is one "super-player" that outperforms all others by far and gets an immense amount of DPoints. They have a perfect score, but as they play only inferior players IMHO they should not be listed in the HoF, and if they get an entry not that high.
Also winning a 34-player game is so much more time consuming and harder than a 2-player duel, esp if your opponents all know how to play.

1v1 excluded :P

Oli - just reporting this as I think my post was drowned...

Does this list update automatically as the games finish? Because gameID=10863 just ended and my Velo didn't change...

No it does not autoupdate. We are still in planning phase...

But if we take into consideration "group of friend" games (perhaps due to language etc) then no system can be ok.

Perhaps making a formula about skilled players according to their wins-draws i.e. those who have 50% win into one category , those who have 40% win+draw get to be in another category of skill

Oli, why don't you hide the list autoupdate?
Looks like most of us is more interested in "personal score" than in how the system works...

Oh one more thing.
Does that mean you will cancel the bet system?

No.
Problem with the Elo-System is: After enough games it will reflect your score pretty well and keep at this level (if you don't improve).
With DPoints a win or draw will always net you some points (as unnecessary as these points are, but player like to "gain" something, even if it's just some strange score in a MySQL database.)

The new rating is just for statistics (the HoF) and will be added after your name (like the DPoints).

I didn't think of the "group of friends" situation.

If an experienced player instruct a bunch of beginners, they could get the highest possible rating after only a few games. Individual players could get out of balance with the rest of the community if they don't play against anyone else.

It might be better to make sure a large enough sample of diverse players are competed against before giving them a rating.

I think this might be part of the phenomenon that is happening with Grace (not the friend part, just the small sample size) who has 1 win and no draws, but is in 14th place. That's a better ranking than some very good players like Jimbozig, Decima, Guaroz, President Eden! All that for one win.

That's why it might be a good idea to award (or display) a rating after 10 games played. Maybe do not award points for the first 10 games for all player involved and once a player hits 10 games create rating based on his last 10 games. Good players can't loose points agains a good player that is wrongly rated with only 1000 , and the players in question would enter the HoF with a more reasonable rating.

It would be ok to award points, but I think 10 games would not be enough. I would propose a minimum of 25 to 50 games required to be displayed as a "rated" player.

Simply not displaying them on the HoF page would be adequate. Their skill as a "non-rated" player could still be displayed on their own page, just not the HoF listing.

Not that it had anything to do with my suggestion of requiring 25-50 games. But, I just noticed Talrus and Orpheus have 24 games finished. :-)

Which means that I'd go up to second :D (Possibly even first considering the game that finished today). I like that idea!

Yeah, I was just looking out for you, Captain. :-)

Requiring that a player has a minimum of 25 to 50 games to be displayed as a "rated" player is, in my opinion, absolutely excessive.

It can take several months to finish a game with a 24H Phase. Many people can't play so many games at a time. I myself have finished "only" 4 games since I joined vdiplomacy a little more than 5 months ago.

Do you realize that a requirement of 25 to 50 games would exclude MANY players, possibly for years ?

For example, in the current rating, among the "top 20 players", 12 of them haven't finished 50 games, and 8 of them haven't finished 25 games !

I wonder what the percentage of players on vdiplomacy who have finished at least 50 games is. Oli do you please have a way to easily know ?

I agree with Retillion. I would say put an asterisk by the rating of players with ten games or less, and then explain the issue of small sample size on the page with the rankings. This would make things clear to all parties.

My opinion is that to be a "rated" player, you should be willing to demonstrate their ability.

The fact almost half of the top 20 have not even played 25 games is an indication that there is a problem with the system. President Eden has a 64% win rate and doesn't make the top 20! That's a problem! Decima has won or drawn 66% of the time, and is in the 45th place. But you think you deserve to be in the top 20 with 2 wins?

Lets not leave out drano with a 90%solo win/draw who is in 61st! :o

:D

No. If the asterisk doesn't make that clear, then just leave out the players with under ten games. At some point though, there will still be players with 125 games being ranked "unfairly" alongside players with 25 games. And the players who play dozens of one vs. one games vs. players like myself, who only play variants with at least five other players.

G-man,

Actually, I was referring to Retillion who also has only 2 wins and 4 games played, but is in 6th place. Your stats are pretty good, but still only have 14 games complete. In my opinion, that is still not enough to be offset a big win.

I agree about playing a bunch of 1v1 games, though. You could change add a minimum # of pairings. You could knock off 34 for a single WWIV game or 33 for a game of Chaos. For example, make it a combination of 20 games with at least 150 pairings. I think that would be a reasonable sample size to say a player is experienced enough to be "rated".

Sorry I missed you Drano. Impressive stats. It seems like a higher ranking would be justified.

Imma pop in with a link and then pop out..

http://www.lifewithalacrity.com/2006/01/ranking_systems.html

Thanks Leif,

If you scroll to the DAYS OF WONDER part, they are tackling the multi-player problem the same way Oli is. They also suggest that anyone playing less than 20 games be considered a "provisional" player.

The highly interesting comment I noted in there was the one of the lines about the Days of Wonder forums.

"There has also been a lot of discussion regarding Ticket to Ride, a strategy game that supports 2-5 people, and whether the ELO variant system discourages multiperson play."

Our variety of multiplayer games would exacerbate any 'discourages multi-person play' issues with our extremely wide range of 2-35 player games.

So the question then becomes how does soloing a 35 player game compare in relation to winning 34 1v1 games? Can they even be compared? It is my opinion that there isn't going to be a feasible direct comparison, and that any ranking system we pick will be at least partially inaccurate and/or subjective especially for people who are outliers (mostly play WWIV/Chaos or mostly play 1v1 for instance).

It is my belief that we will have a sizable and vocal minority that is extremely dissatisfied with whatever ranking system we implement. It is for this reason that I propose a plurality of ranking systems, and that most importantly we provide per variant and/or 'per number of player' rankings (within which we can much more definitely make comparisons).

Ghost rating works well for webDip in that they predominately rank 1 type of variant (ancient med is not counted right? what about the 17 player world variant?), but I'm not certain that it will work well here.

I'd love to see on the variant description page a stat that shows the top 20 players of that variant.

And I also agree that a plurality of ranking systems might be the best way to provide information, and for making comparisons.

+1 Leif

Good point Leif. Apparently there is room to have several columns on the HoF listing. Why not include several different ranking systems?

Pardon,
I’ve done the wrong hand calculation. Correcting my earlier post data:

Using the current Elo-based formulation, by a player’s point of view:
After a duel my rating can change +/- 15
After a classic my rating can change +/- 15*6 = +/- 90
After a KW_901 my rating can change +/- 15*14 = +/- 210
After a WWIV my rating can change +/- 15*34 = +/- 510

I remind that using the coefficient of 400 in the exponential model means this:
+400_WDL : ten times more skilled
+191_WDL : three times more skilled
+120_WDL : twice more skilled
+70_WDL : 50% more skilled
+32_WDL : 20% more skilled
+15_WDL : 9% more skilled
+0_WDL : equally skilled

Using this system the DIFFERENCE of WDL points is the only significant measure. The WDL difference is taken into account when evaluating the expected performances. Other measures (he has twice my points, he is 14th and I am 19th …) are useless and/or misleading.

This said, no system can reliably rate an user with few data (few games played). This happens with the current D-point system, with the RR-system… There’s no way around this statement. It’s up to our common sense understanding when the user rating makes sense or not. There’s no reason to exclude users from the list.

What happened?
Did I switch off the discussion?